Skip to Main Content

Topic: Systematic Searching for Evidence Synthesis

What is PubMed?

PubMed is a free database provided by the US government. It primarily consists of MEDLINE, a dataset of over 36 million citations from the biomedical literature.

If you've searched MEDLINE on other platforms (like EBSCOhost or Ovid), the difference in content is that PubMed also includes PubMed Central, an archive of full text journal articles; and the Bookshelf, an archive of books, reports, and other materials. In terms of searching, the technical syntax in PubMed is very different from what you would use on those other platforms, even though the controlled vocabulary is the same.

How Does PubMed Search?

PubMed's default search uses Automatic Term Mapping (ATM) to make an educated guess as to what you're looking for. The most important thing ATM checks for is whether there is a good match for the words or phrases in your search within PubMed's controlled vocabulary (MeSH, or Medical Subject Headings).

ATM is fine, but in a systematic search, you'll need to have full control of what PubMed is doing. That means:

  • Using the Advanced Search page
  • Identifying appropriate controlled vocabulary
  • Constructing your own complex search strings and combining them thoughtfully
  • Using search tags to tell PubMed where to search for terminology

Searching PubMed's Controlled Vocabulary: MeSH

Controlled vocabulary is the specialized language used by a database to describe the citations within it. The controlled vocabulary used in PubMed is called MeSH, or Medical Subject Headings. Below is a video about two ways to locate relevant MeSH headings and use them to further develop your concept table.

Building a Systematic Search in PubMed

Watch the video below to learn how to build a search strategy in PubMed using keywords and controlled vocabulary, AND and OR, and PubMed's unique search syntax.

It's fairly long, but it is essential to understanding the process. Don't forget you can download the transcript to use in conjunction with the video or to refresh your memory of the steps as you construct your own search. Further down this page, you'll find a reference box that provides text examples of important elements covered in this video, such as search tags, truncation, and phrase searching.

Now that you've watched the video, you're ready to begin your exploratory searches in PubMed.

For Your Reference: Search Tips for PubMed

PubMed is a database, which means each reference is a record made up of many different fields. Some examples of fields include title, abstract, and MeSH. Search tags allow you to tell PubMed where in the record to search. As you construct your search, you'll typically be looking in a few specific fields:

[mesh]

Searches only the MeSH field.

An important note: In general, using a search tag like [mesh] both requires PubMed to search in that specific field and prevents it from processing your words through Automatic Term Mapping (ATM). In rare cases, when the word or phrase you use appears in more than one MeSH term (such as drinking and alcohol drinking), PubMed will map to any MeSH term containing your word or phrase. In this example, if you search drinking [mesh] without quotation marks, PubMed will search for drinking [mesh] OR alcohol drinking [mesh]. Searching "drinking" [mesh] prevents the unintended mapping.

Because of this possibility, it's good practice to use "quotation marks" around words or phrases when using the [mesh] search tag.

[mesh:noexp]

Searches only in MeSH, and explicitly excludes the narrower terms underneath your MeSH term in the hierarchy (turns off "exploding").

[tiab]

Searches in title, abstract, and keywords (as well as collection title and other abstract).

[tiab] is the best choice, in most cases, for searching keywords from your concept table. You should also use it to search your MeSH terms as keywords, just in case!

[tw]

This searches in lots of different fields, including title, abstract, MeSH, MeSH subheadings, publication types, substance names, various author name fields, and more.

Because [tw] includes the MeSH field, it has a tendency to duplicate your MeSH search. It may also be too sensitive, returning lots of irrelevant results ("noise"). I recommend using [tiab] instead, in most circumstances. There may be cases in which [tw] is appropriate, but we can work that out together.

In databases, searching for any ending of a word is called truncation. Frequently you can use * (asterisk) to truncate a word, but the symbol can vary across different databases. In some cases, you can use a wildcard to find alternative spellings that fall inside a word, rather than at the end.

Essential things to know about truncation in PubMed:

  • The * in PubMed substitutes for 0 or more characters, either within (wildcard) or at the end (truncation) of a word, or even both within and at the end of the same word.
  • There must be at least 4 characters before the first use of the * wherever you put it: organi*ation* will work, but org*ation* will not, nor will run*.
  • Truncation disables ATM, which is fine, since you'll generally be avoiding using ATM in your searching.
  • You can truncate any or all words in a phrase. To do this:
    • Use double quotes and a search tag: "cardio* exercis*" [tiab] (this is what I generally recommend).
    • Use double quotes around the phrase: "cardio* exercis*".
    • Use a hyphen: cardiovascular-exercis*.

Phrase searching works differently in PubMed than in most databases, so see the next tab in this box to learn more about it.

There are three different ways to do phrase searching for keywords in PubMed:

Option 1 (this is what I recommend)

Use a [search tag]: cardiovascular exercise [tiab]

PubMed first checks to see if the phrase appears in its phrase index. If it does, PubMed executes the search as requested. If it doesn't, PubMed ignores the search tag and tries to match the words x AND y using ATM.

Option 2

Put "double quotes" around your phrase: "cardiovascular exercise"

PubMed first checks to see if the phrase appears in its phrase index. If it does, PubMed executes the search as requested. If it doesn't, PubMed tries to match the words x AND y using ATM.

Option 3

Use a hyphen: cardiovascular-exercise

PubMed first checks the phrase index. If it isn't there, PubMed returns no results for your phrase.

Proximity searching in PubMed can be used in very specific cases.

  • When searching for terms appearing in any order within a specified distance of one another.
  • Only when searching in the title, abstract, or affiliation fields; this means you can only use it with the following search tags: [ti] [tiab] [ad].
  • Wildcards and truncation cannot be used in proximity searches.

The syntax for proximity searching is "search terms" [field:~N] where N is the number of words between terms.

Proximity searching can be tricky in PubMed. I recommend the resources below for further details and explanations.

Using Search Hedges in PubMed

Search hedges, sometimes called search filters, are designed to narrow your results based on specific criteria, such as publication type. They are designed by experts to be as comprehensive as possible.

Watch the video below to learn how to use a search hedge, specifically the CADTH hedge which limits to systematic reviews and other evidence syntheses.

You can find the full hedge to copy and paste into your PubMed search in the next box on this page. Don't forget to acknowledge and cite any hedges you use in your searches!

Common Search Hedges & How to Find Others

Below is the CADTH search hedge to limit to systematic reviews and other evidence syntheses in PubMed. To use this, as with most search hedges, follow these steps:

  1. Do a search for the full text of the hedge in the PubMed Advanced Search query box.
  2. Use AND to combine the results of the search hedge with the search or searches that reflect the concepts in your research question.

You can copy and paste the text below to execute the first step above.

"systematic"[filter] OR "meta-analysis"[pt] OR "meta-analysis as topic"[mh] OR "meta analy*"[tw] OR metanaly*[tw] OR metaanaly*[tw] OR "met analy*"[tw] OR "integrative research"[tiab] OR "integrative review*"[tiab] OR "integrative overview*"[tiab] OR "research integration*"[tiab] OR "research overview*"[tiab] OR "collaborative review*"[tiab] OR "collaborative overview*"[tiab] OR "systematic review"[pt] OR "systematic reviews as topic"[mh] OR "systematic review*"[tiab] OR "technology assessment*"[tiab] OR "technology overview*"[tiab] OR "technology appraisal*"[tiab] OR "Technology Assessment, Biomedical"[mh] OR HTA[tiab] OR HTAs[tiab] OR "comparative efficacy"[tiab] OR "comparative effectiveness"[tiab] OR "outcomes research"[tiab] OR "indirect comparison*"[tiab] OR "Bayesian comparison"[tiab] OR (("indirect treatment"[tiab] OR "mixed-treatment"[tiab]) AND comparison*[tiab]) OR Embase*[tiab] OR Cinahl*[tiab] OR "systematic overview*"[tiab] OR "methodological overview*"[tiab] OR "methodologic overview*"[tiab] OR "methodological review*"[tiab] OR "methodologic review*"[tiab] OR "quantitative review*"[tiab] OR "quantitative overview*"[tiab] OR "quantitative synthes*"[tiab] OR "pooled analy*"[tiab] OR Cochrane[tiab] OR Medline[tiab] OR Pubmed[tiab] OR Medlars[tiab] OR handsearch*[tiab] OR "hand search*"[tiab] OR "meta-regression*"[tiab] OR metaregression*[tiab] OR "data synthes*"[tiab] OR "data extraction"[tiab] OR "data abstraction*"[tiab] OR "mantel haenszel"[tiab] OR peto[tiab] OR "der-simonian"[tiab] OR dersimonian[tiab] OR "fixed effect*"[tiab] OR "multiple treatment comparison"[tiab] OR "mixed treatment meta-analys*"[tiab] OR "umbrella review*"[tiab] OR (("multiple paramet*"[tiab]) AND ("evidence synthesis"[tiab])) OR (("multi-paramet*"[tiab]) AND ("evidence synthesis"[tiab])) OR ((multiparameter*[tiab]) AND ("evidence synthesis"[tiab])) OR "Cochrane Database Syst Rev"[Journal] OR "health technology assessment winchester, england"[Journal] OR "Evid Rep Technol Assess (Full Rep)"[Journal] OR "Evid Rep Technol Assess (Summ)"[Journal] OR "Int J Technol Assess Health Care"[Journal] OR "GMS Health Technol Assess"[Journal] OR "Health Technol Assess (Rockv)"[Journal] OR "Health Technol Assess Rep"[Journal]

In future searches, you may need additional search hedges. These are my most-used sources of hedges, although they're certainly not the only place to find them.

Exporting Results from PubMed

To export all results from a PubMed search, up to 10,000 citations, choose the Send to button above your results. Select Citation Manager from the list of options.

In the Selection drop-down, choose All results.

Tap the Create file button. This will download a file in nbib format, which can be imported into the citation management tool of your choice.

Downloading Your PubMed Search History

Most databases allow you to download your search history. To do so in PubMed, go to the Advanced Search page and tap the Download button at the top right of the History and Search Details box.

History and Search Details box, with an arrow pointing to the Download link.

This downloads your search history as a csv file. When you open it in Excel, you can expand the cell sizes and wrap the text to see your search strings more clearly. Because Excel has a character limit for each cell, you may need to open your csv file in a text editor like Notepad to see the full text of very long search strings.

Downloading your searches allows you to easily see and document how you searched as well as to re-execute searches by copying and pasting search strings.

Creative Commons  This work is licensed under a Creative Commons Attribution NonCommercial 4.0 International License. | Details and Exceptions